paper: tighten Software Design prose to keep Table 1 on one page

Trim ~18 words of filler from the Package design and Software architecture
paragraphs so Table 1 no longer breaks across pages 2-3.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: pavelkomarov

paper/paper.md

@@ -59,9 +59,9 @@ The original PyNumDiff publication [@vanBreugel2022] established the core method
 x_hat, dxdt_hat = method(x, dt_or_t, **params)
 ```
 
-where `x` is a NumPy array [@harris2020array] of measurements; `dt_or_t` is either a scalar step size or an array of sample locations; and keyword arguments configure the method. Explicit keyword arguments make calls self-documenting; prior positional signatures are preserved with deprecation warnings.
+where `x` is a NumPy array [@harris2020array] of measurements; `dt_or_t` is either a scalar step size or an array of sample locations; and keyword arguments configure the method, making calls self-documenting. Prior positional signatures are preserved with deprecation warnings.
 
-**Software architecture.** PyNumDiff is organized into seven method modules plus shared `utils` and `optimize` modules, a flat structure chosen for discoverability. Where strong alternatives exist, PyNumDiff delegates rather than reimplements: SciPy [@virtanen2020scipy] provides spline fitting, Savitzky-Golay filtering, and signal processing routines; NumPy [@harris2020array] provides the FFT; PyWavelets [@lee2019pywavelets] provides the discrete wavelet transform for `waveletdiff`; CVXPY [@diamond2016cvxpy] handles convex optimization for `robustdiff` and `tvrdiff`, as an optional dependency, keeping the base installation lightweight. The `kalman_filter` and `rts_smooth` primitives are public, letting users with known dynamical models bypass the assumed constant-derivative model of `rtsdiff`; an `innovation_fn` hook extends the filter to non-Euclidean spaces.
+**Software architecture.** PyNumDiff is organized into seven method modules plus shared `utils` and `optimize` modules in a flat structure. Where strong alternatives exist, PyNumDiff delegates rather than reimplements: SciPy [@virtanen2020scipy] provides spline fitting, Savitzky-Golay filtering, and signal processing routines; NumPy [@harris2020array] provides the FFT; PyWavelets [@lee2019pywavelets] provides the discrete wavelet transform for `waveletdiff`; CVXPY [@diamond2016cvxpy] handles convex optimization for `robustdiff` and `tvrdiff` as an optional dependency. The public `kalman_filter` and `rts_smooth` primitives let users with known dynamics bypass `rtsdiff`'s constant-derivative model; an `innovation_fn` hook extends the filter to non-Euclidean spaces.
 
 **Method capabilities.** All non-deprecated methods support multidimensional data via `axis`; Table 1 lists additional specialized capabilities.